pan india opinion poll with detailed regional break up

by truthbetold Sat Apr 05, 2014 4:10 pm

http://www.ibtl.in/news/exclusive/2128/ibtl-public-opinion-poll-2014

A modi sweep is predicted.

Even Cd would have a problem rubbishing this survey since it put 's jagan at or above tdp.

I do not know much about ibtl but it provides lots of statewise detail.

by truthbetold Sat Apr 05, 2014 4:30 pm

http://m.ndtv.com/elections/article/election-2014/despite-local-opposition-bjp-set-to-finalise-deal-with-chandrababu-naidu-in-andhra-pradesh-504868

Look like this long twisted tdp bjp alliance got finalized. Seemandhra both sides may benefit. T area it is a disaster similar to 2009 deal with trs.

by rawemotions Sat Apr 05, 2014 10:07 pm

truthbetold wrote:http://www.ibtl.in/news/exclusive/2128/ibtl-public-opinion-poll-2014

A modi sweep is predicted.

Even Cd would have a problem rubbishing this survey since it put 's jagan at or above tdp.

I do not know much about ibtl but it provides lots of statewise detail.

NDTV poll gave sample size, methodology etc.. I do not see the same here.

Unless the sample size > 1000 per constituency, it cannot be trusted. That means it should be atleast 5.5 L. Even NDTV had just 2 L.

Even then it is possible to skew an opinion poll. A "Yes Minister" serial once explained in a detailed manner how we can skew any sort of opinion poll with intelligent choice of questions.

by confuzzled dude Sat Apr 05, 2014 10:15 pm

truthbetold wrote:http://www.ibtl.in/news/exclusive/2128/ibtl-public-opinion-poll-2014

A modi sweep is predicted.

Even Cd would have a problem rubbishing this survey since it put 's jagan at or above tdp.

I do not know much about ibtl but it provides lots of statewise detail.

"In his life, Chandra Babu Naidu has never won an election on his own. He took over as the CM of Andhra from NTR in mid 90s. In 1998 TDP could just win 12 out of 42 seats. BJP managed to win 4 seats on its own and grabbed a voteshare of 18%. In 1999 Naidu sensed the opportunity and sensed that an alliance was beneficial to the TDP. The 18% votes of BJP helped TDP win a whopping 29 seats in 1999 aong with the countrywide Vajpayee wave. BJP also settled for a meagre 8 seats and won 7 out of the 8."

-> Wow! and certain folks blame me as being prejudiced when I say the same.

by rawemotions Sat Apr 05, 2014 10:43 pm

Same with YSR he was down in the dumps and he rekindled the Telangana problem by inciting TRS and then made an alliance with them to take on TDP.

by confuzzled dude Sat Apr 05, 2014 10:48 pm

rawemotions wrote:Same with YSR he was down in the dumps and he rekindled the Telangana problem by inciting TRS and then made an alliance with them to take on TDP.

He won in '09 on his own despite corruption talk & CBN lost in '04 despite development talk. See the difference.

by truthbetold Sat Apr 05, 2014 11:11 pm

Raw,

I do not much about IBTL. I said that earlier.

How did you come with 1000 per constituency figure?

A general opinion poll is conducted in a sample of constituencies from a sample from each constituency. The locations are selected based on past results and locations ability to reflect local, regional and national trends. Sample of people is selected to get a representative mix of urban /rural, religion, economic class and caste groups.

USA voter stratification and information sources are different. Here the sample is much smaller. That is why I am curious to know how you came up with 1000. Note that larger sample size does not improve poll precision in proportion to sample size. Poll methodology and execution are more important once a scientific sample size is determined.

by Marathadi-Saamiyaar Sat Apr 05, 2014 11:15 pm

rawemotions wrote:
truthbetold wrote:http://www.ibtl.in/news/exclusive/2128/ibtl-public-opinion-poll-2014

A modi sweep is predicted.

Even Cd would have a problem rubbishing this survey since it put 's jagan at or above tdp.

I do not know much about ibtl but it provides lots of statewise detail.
NDTV poll gave sample size, methodology etc.. I do not see the same here.

Unless the sample size > 1000 per constituency, it cannot be trusted. That means it should be atleast 5.5 L. Even NDTV had just 2 L.

Even then it is possible to skew an opinion poll. A "Yes Minister" serial once explained in a detailed manner how we can skew any sort of opinion poll with intelligent choice of questions.

This is very well known. Statistics is a valuable tool only if the analysis is carried out in an unbiased manner by a non-interested party. That is a big dream in India.

and the sheer size of each constituency and the vast difference among the population in each constituency is a real statistical challenge. Even if the final results actually match any of the predictions it is only by coincidence.

by rawemotions Sun Apr 06, 2014 1:58 am

truthbetold wrote:Raw,

I do not much about IBTL. I said that earlier.

How did you come with 1000 per constituency figure?

A general opinion poll is conducted in a sample of constituencies from a sample from each constituency. The locations are selected based on past results and locations ability to reflect local, regional and national trends. Sample of people is selected to get a representative mix of urban /rural, religion, economic class and caste groups.

USA voter stratification and information sources are different. Here the sample is much smaller. That is why I am curious to know how you came up with 1000. Note that larger sample size does not improve poll precision in proportion to sample size. Poll methodology
and execution are more important once a scientific sample size is determined.

The goal is to have tight confidence intervals , and essentially 2-3% error. I feel 4%-5% reported by current pollsters is too large for a country like India to use as an indicator to predict seats.

[size=12.800000190734863]This is hand-waving and not based on any historical data. I agree a good number might need to include all those parameters you talk about.[/size]

Thus, It was a guess estimate. In any constituency we want to reach all the communities/sub-cultures. The smallest communities might be only 10% of the population in a constituency. Now I just made a guess that we need at least 100 persons (50 men/50 women)from the smallest constituency to get an accurate picture of what that community thinks, if I need to have 100 persons for the smallest community/sub-culture with 10% population. Then, as you said I have various classes urban/rural/ and other sub cultures (3-4) and put together, I have 10 different groups of people to cover, 100 each. Some constituencies are large so we need to cover multiple areas spread geographically. The number 1000 becomes higher and may be we need 3000 in such places.

There might be small constituencies where only two communities are present in which case, make do with may be just 500. But given the diversity in India and demographics I am guessing that, the required confidence intervals (2-3% error, not 4-5% claimed by current pollsters) will be met statistically with an approximate number of 1000 per constituency.

by Rishi Sun Apr 06, 2014 5:14 am

23 seats by Congress in TN! No way.

by Rishi Sun Apr 06, 2014 6:16 am

My opinion.

Modi Should shut up. Must learn from Putin. Otherwise, he will not last even a month.

by truthbetold Sun Apr 06, 2014 6:28 am

Rishi wrote:23 seats by Congress in TN! No way.

Rishiji
23 in south, 0 in tn

by rawemotions Sun Apr 06, 2014 8:28 am

Rishi wrote:My opinion.

Modi Should shut up. Must learn from Putin. Otherwise, he will not last even a month.

Good suggestion!

by truthbetold Sun Apr 06, 2014 9:14 am

Raw,

The purpose of the opinion polls defines how to select the sample. If news media conducts opinion polls to project party positions in a state, they can focus on who will ahead of others in a constituency. Precision in terms of which caste is voting for which candidate is secondary or not important. on the other hand, a BJP conducted internal poll for a specific candidate in a specific constituency would be more interested in swing voter preferences and their reasoning. Then party will be able identify actionable groups and messages. All parties do this using experienced pollsters.

Newsmedia opinion polls should be understood as directional indicators and a rough indicators of distances between different parties. They were never intended to be precise seat count mechanisms as some project them to be.

by truthbetold Sun Apr 06, 2014 9:14 am

Raw,

The purpose of the opinion polls defines how to select the sample. If news media conducts an opinion polls to project party positions in a state, they can focus on who will be ahead of others in a constituency. Precision in terms of which caste is voting for which candidate is secondary or not important. On the other hand, a BJP conducted internal poll for a specific candidate in a specific constituency would be more interested in swing voter(caste group) preference and its reasoning. Then party will be able identify actionable groups and messages. All parties do this using experienced pollsters. Not all the money goes to liquor.

Newsmedia opinion polls should be understood as directional indicators and a rough indicators of distances between different parties. They were never intended to be precise seat count mechanisms as some project them to be.

by MaxEntropy_Man Sun Apr 06, 2014 10:01 am

truthbetold wrote:
Rishi wrote:23 seats by Congress in TN! No way.
Rishiji
23 in south, 0 in tn

why not rishi-gAru or rishi-ayyA?

by truthbetold Sun Apr 06, 2014 10:09 am

MaxEntropy_Man wrote:
truthbetold wrote:
Rishi wrote:23 seats by Congress in TN! No way.
Rishiji
23 in south, 0 in tn

why not rishi-gAru or rishi-ayyA?

why not rishiji?

by MaxEntropy_Man Sun Apr 06, 2014 10:11 am

truthbetold wrote:
MaxEntropy_Man wrote:
truthbetold wrote:
Rishi wrote:23 seats by Congress in TN! No way.
Rishiji
23 in south, 0 in tn

why not rishi-gAru or rishi-ayyA?
why not rishiji?

i find it odd that you choose to add a hindi suffix, when one is more readily available in your native tongue or the native tongue of the person being addressed.

by truthbetold Sun Apr 06, 2014 10:34 am

MaxEntropy_Man wrote:
truthbetold wrote:
MaxEntropy_Man wrote:
truthbetold wrote:
Rishi wrote:23 seats by Congress in TN! No way.
Rishiji
23 in south, 0 in tn

why not rishi-gAru or rishi-ayyA?
why not rishiji?

i find it odd that you choose to add a hindi suffix, when one is more readily available in your native tongue or the native tongue of the person being addressed.

Hindi/urdu is one of my communication tools. I use hindi/english/telugu without consciously selecting between them.

by Sponsored content

pan india opinion poll with detailed regional break up

pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up

Re: pan india opinion poll with detailed regional break up